NHS R Conference

This is a subtitle

Dec 9, 2024

The conference

Some information

The Talks

Day 1
Day 2

Time	Title	Speaker
09:45 – 09:55	Welcome and Housekeeping	Zoë Turner
09:55 – 10:20	RAPping my head against a databricks wall	Louise Schreuders
10:20 – 10:35	Reducing mental health inequalities among the BAME residents of Herefordshire and Worcestershire communities	Oluwatimilehin Olabamiyo
10:35 – 11:00	Simplifying development of AI applications integrated in EHRs with Health Chain	Jennifer Jiang-Kells
11:00 – 11:15	Beyond the dashboard: R for value added insights	Nicola Farthing
11:45 – 12:00	Reimagining NHS dashboards: an open-source approach with plotly-dash	Jennifer Struthers
12:00 – 12:25	Using Machine Learning and secondary care activity data to identify risk of cancer earlier	Scarlett Kynoch
12:25 – 12:40	The patient does not exist – generating synthetic patient data with Wasserstein GAN	Simon Newey
12:40 – 13:05	Streamlining machine learning development at the NHS via open-source tools	Elias Altrabsheh and James Sibbit
13:05 – 13:07	rainbowR	Ella Kaye
13:55 – 14:10	What insights did Glasgow Scottish Ambulance Service (SAS) gain from combining multiple data sources about all chest pain patients from 2023? We'll present about the process and findings of a 1-year long MSc dissertation project.	Katalin Koszegi
14:10 – 14:25	Predictive Modelling for health and social care capacity planning using open data	Sebastian Fox
14:25 – 14:50	To explain or predict: how different modelling objectives change how you use the same tools	Chris Mainey
14:50 – 15:05	Using Openxlsx2 to automate excel publications	Ruth Keane
15:35 – 15:50	What I learnt about (programming) languages by building bilingual websites	Rosemary Walmsley
15:50 – 16:15	Leveraging R to implement novel theoretical development in online ‘digital twin’ simulation modelling	Richard Wood
16:15 – 16:30	Should I use your package	Colin Gillespie
16:30 – 16:45	Cracking open the TiN: how we build a one-stop statistics website using R, GitHub and BigQuery	Mohan Del

Time	Title	Speaker
09:45 – 09:55	Welcome and Housekeeping	Zoë Turner
09:55 – 10:20	The Reusability Crisis in Healthcare Analytics	Rhian Davies
10:20 – 10:35	Shift staffing via task load prediction	Marcos Fabietti
10:35 – 10:50	Unleashing the power of pathway simulation	Sammi Rosser
11:15 – 11:40	New generic tests for cancer – with R is a clinical scientists best friend	Joe Shaw
11:40 – 11:55	Beyond automation: a shiny app to maximise analytical impact routine reporting narrative	Laura Birks
11:55 – 12:10	Sharpening my Python skills through self-development of web scraping bank complaints data	Kenneth Quan
12:10 – 12:25	GitHub as a team sport	Matt Dray
13:10 – 13:25	Presenting fingertips in data in a more friendly format	Rachel Brown
13:25 – 13:50	A method to apply temporal graph analysis on electronic patient record data to explore healthcare professional patient interaction intensity	John Booth
13:50 – 14:05	Deploying a Shiny app with Docker in a Raspberry Pi	Pablo León Ródenas
14:05 – 14:20	Estimating flexible hazard rates for C diff recurrence from electronic health records using the SplinHazard Regression package and other methods in R	Elisabeth Dietz
14:20 – 14:45	Assessment of patient feedback using Natural Language Processing (NLP) and textual data analysis in R	Ana Singh
15:15 – 15:40	Forged in the fire: agile project management lessons from the frontline	Chris Beeley
15:40 – 15:55	Community Talk – Turing Way	Sophia Batchelor
15:55 – 16:05	Community Talk -NHS.Pycom	Alex Cheung
16:05 – 16:15	Closing talk NHS-R Community and raffle	Zoë Turner

RAP

wRAPping your head around the idea

RAP stands for

Reproducible Analytical Pipeline

Aim: create reproducible code to streamline a repetitive process such as regular reports or statsitcal analysis.

wRAPping your head around the idea

Key ideas of RAP:

Automation
Modular, re-usable code
Transparency
Open Standards and Tools
Version Control

RAPs are automated analytical processes including everything from initial data extract, data cleaning, analysis and modelling to visualisation and reporting. What’s special about a RAP is that it’s designed to ensure reproducibility, quality and efficiency.

RAP principals:

Automation:
- Faster processes: Much faster to get the computer to do all the work!
- More resource: Free up analyst time to work on other things
- Accuracy: Reduces the risk of human error
Modular, re-usable code
- Avoid repetition: For example if you need to suppress counts in multiple places in your script, you can just call the function each time - no need to copy and paste the same code in several places
- Re-usability: Easier to re-use code in other work, speeding up the development of future projects
- Integration: Easier to “slot in” other people’s functions - no need to reinvent the wheel if a solution already exists!
- Faster on-boarding: Code is easier to understand, so new users can get up to speed faster
Transparency:
- Improve public trust: through transparency of our processes
- Identify mistakes: More eyes on the code makes it more likely that mistakes will be spotted
- Feedback: Readers of the code are able to suggest and make improvements to it
- Collaboration: Increases collaboration and knowledge sharing
Open source
- Free: open-source tools are completely free to use
- Support: they tend to have a very active community of helpful people to seek advice and support from
- Reusability: anyone can run open-source code, you don’t need to be a fellow user of a proprietary system. This makes it easier to share work across teams and organisations
- Use other people’s code: Reusability goes both ways! We can take advantage of the libraries, packages, and other code made by others
Version Control:
- Auditability: Create an audit trail of changes made, logging what changes were made when and by whom
- Roll backs: Enable users to revert changes, or roll back to previous versions
- Enhanced collaboration: Makes collaboration on coding projects much easier
- Code quality: Makes it easier to peer-review code for quality and reliability
- Transparency: Enables code to be shared more easily, using cloud services like GitHub or GitLab.

wRAPping your head around the idea

Key ideas of RAP:

Automation
Modular, re-usable code
Transparency
Open Standards and Tools
Version Control

RAPping your head around the idea

flowchart LR
  A[Data Store] --> B(SAS)
  B --> |Sometimes automated?| C(Word)
  B -.-> D(Excel?) -.-> C
  C --> E(PDF)
  F{Quality Assurance} --> B 
  F --> C
  E --> G[Published]

  subgraph Loop["Repeat every time the data / analysis changes"]
    B
    C
    D
    F
  end
  
  style Loop fill:none,stroke:#000,stroke-width:2px;

The non-RAP way Let’s compare this to the current state of play that we are trying to move beyond with RAP. Imagine a publication process that doesn’t use the principles above. It might involve:

Manually running queries against a data store to retrieve, join, and clean the data
Exporting the data and manually importing it into another statistical or analysis package
Creating some statistics/outputs
Manually copying these to an Excel/Word/PDF report

The gist is: a processes requiring a lot of manual steps, and usually using proprietary software.

This has several downsides such as:

Time consuming: manual steps are time consuming, not only to carry out, but because they introduce a point where further manual quality assurance (QA) is needed
Potential for error: Fiddly manual steps introduce the possibility of human error. Lack of tests mean we can’t be completely sure how future changes affect the old code
Scalability: If you need to add a new table or statistic to the publication, the process will take longer to complete
Difficult to extend: Code written on in a non-modular way using proprietary systems is harder to extend when changes need to be made

But why bother?

The Sudlow Review: ‘Uniting the UK’s Health Data: A Huge Opportunity for Society’
an independent review commissioned by the Chief Medical Officer for England, the UK National Statistician and NHS England’s National Director for Transformation.

Stakeholders’ highest priorities for improvements…in the health data ecosystem…
- Improve data usability: the implementation of open, shareable and reproducible approaches to data management, curation and analysis pipelines will drive transparency and efficiency, and reduce duplication of effort

But why bother?

Better, broader, safer: using health data for research and analysis
review into how the efficient and safe use of health data for research and analysis can benefit patients.

Summary recommendations
7. Promote and resource ‘Reproducible Analytical Pathways’ as the minimum standard for academic and NHS data analysis: this will produce high quality, shared, reviewable, re-usable, well-documented code for data curation and analysis; minimise inefficient duplication; avoid unverifiable ‘black box’ analyses; and make each new analysis faster.

Some reading

NHS Digital RAP website, including information on different ‘RAP levels’
NHS England blog post
Government Analysis Function RAP strategy
Government Data Science RAP Companion
Office for Statistic Regulation: Overcoming barriers to adoption
RAP with R book
DfT RAP Implementation Plan